Previous research comparing the reinforcing efficacy of tokens and primary reinforcers produced potentially confounded findings. Bonfonte, Bourret, and Lloveras (2020) argued this confounding nature of the token conditions of previous studies, citing that the participants had access to primary reinforcers during conditions designed to assess the reinforcing efficacy of tokens. Consequently, it was unclear whether participants’ responses were maintained by the tokens themselves, by the primary reinforcers, or by a combination of both. Thus, the authors of this study examined the reinforcing efficacy of conditioned reinforcers relative to primary reinforcers (like food) by isolating the token condition from the primary reinforcer conditions. They achieved this by employing a translational approach. A translational approach involves the use of experimental methods from basic research to explore real-world interventions, such as token economies. The goal was to replicate intervention conditions on a smaller scale to study the efficacy of the conditioned reinforcer. The authors argued that doing so could help isolate and highlight the stimulus function of the tokens as conditioned reinforcers.

​

Bonfonte et al. (2020) reported on the frequent use of token economies as interventions in applied behaviour analysis across various settings, including educational as well as clinical settings. Token economies entail the accumulation and eventual exchange of previously neutral stimuli for other, more desired, backup reinforcers. Although tokens are delivered immediately, their exchange for desired items is often delayed. Thus, tokens are relied upon to “bridge the gap” between behaviour’s occurrence and the later reward. However, if tokens are far less desirable than primary reinforcers, they may be less effective when delays become longer or when a large amount of effort is needed to acquire them. 

The present study involved two adolescents with Autism Spectrum Disorder: Walter, aged 16, and Jeffrey, aged 15. Their developmental and communication characteristics differed, which may be relevant when interpreting their individual response patterns.. Jeffrey used laminated tokens and a token board while Walter’s token board had gridlines where a token would be delivered as an “X” using a dry-erase marker. Neither participant had a prior history with the tokens used in the experiment, allowing the researchers to establish the novel tokens as conditioned reinforcers without concerns for previous carryover effects. Sessions were conducted individually in a quiet room, and the target response consisted of a simple card-touching task. Walter alternated between touching two coloured cards, whereas Jeffrey moved a weighted object between two cards.

​

The procedure consisted of four major phases. First, the authors conducted paired-stimulus preference assessments to identify each participant’s most and least preferred edible items. The highest-preference and the lowest-preference items were those at the top and the bottom of the preferential hierarchy while also being reliably consumed. The highest-preference items were Cheez-its and Cheetos while the lowest-preference items were croutons and goldfish crackers for Walter and Jeffrey, respectively. All token exchanges were arranged on a 1:1 ratio, meaning that each token could be exchanged for one edible item from the identified array of preferred items.

​

Upon the conclusion of the preference assessment, the authors established the novel tokens as conditioned reinforcers by repeatedly pairing them with the edible reinforcers. Participants earned tokens by completing the card-touching response, and each earned token was exchangeable during the session. Here, the authors used a second-order fixed-ratio schedule in which tokens could be earned by responding a set number of times and a pre-determined amount of tokens had to be earned before they could be exchanged. Initially, one response produced one token and one token was needed to exchange for one edible item, illustrated as FR 1 (FR 1) where the token-producing responses were depicted within the parentheses while the exchangeable value of tokens was depicted outside the parentheses. The response and exchange requirements were gradually increased once responding remained stable across multiple sessions. This ensured that participants learned the relation between responding, token production, and subsequent access to food, thereby allowing the tokens to acquire the desired conditioned reinforcing properties.

​

The third phase involved a conditioned-reinforcer assessment using a tandem control procedure. This assessment was designed to determine if the tokens themselves were independently reinforcing. The first of the two conditions was the token condition wherein every five card touches produced a token, but no edible reinforcers were visible during the session. Tokens were accumulated throughout the session, and after a one-minute delay, participants exchanged each token for an edible item. The tandem condition was identical to the token condition except for the absence of token delivery; participants completed the same tasks, experienced the same one-minute delay, and received the edible items without the need for exchanging tokens. Because food was available in both conditions, differential responding across the two conditions could be attributed to the conditioned reinforcement produced by the tokens.

​

Lastly, the fourth phase consisted of a progressive-ratio (PR) assessment that compared the reinforcing efficacy of tokens to that of the high- and low-preference edible reinforcers. A progressive-ratio schedule would increase the required number of responses to receive a reinforcer after each reinforcer delivery. This procedure allowed researchers to assess the persistence of responding as the effort required to obtain a reinforcer increased. In this study, the progressive-ratio schedule began at two responses per reinforcer (FR 2) and doubled in each subsequent step (FR 2, 4, 8, 16, and so on). This phase also involved multiple conditions based on the reinforcer delivered (high-preference and low-preference edible condition, as well as a token condition). During token condition, only tokens were delivered as reinforcers and exchanges for edible items occurred after the session following a one-minute delay. This ensured that responses during the session were maintained solely by the token’s deliverance.

 

The authors reported that they had successfully created the conditions for the tokens to function as conditioned reinforcers for both participants. Results from the conditioned-reinforcer test suggested that the token condition produced higher response rates relative to the tandem conditions, which produced substantially low or no responses. The progressive-ratio assessment indicated that high-preference edible reinforcers maintained higher response frequencies than tokens. The relative effectiveness of tokens and low-preference edible reinforcers should be described according to the individual participants' results. Overall, the findings indicated that tokens could maintain responding but did not demonstrate the same reinforcing efficacy as highly preferred edible reinforcers under the conditions examined.

​

Based on the results, the authors concluded that their findings differed from earlier studies primarily because of methodological changes. By separating token delivery from direct access to primary reinforcers during the assessment, the authors addressed a methodological limitation in previous literature. Despite its merits, the authors also reported on the study’s limitations. The low sample size limited the extent to which the study's findings could be generalized to various populations and settings. In addition, the target response of touching coloured cards was relatively simple and likely did not represent complex behaviours that are usually addressed in prevalent token economies. The 1:1 exchange ratio may also limit generalizability to prevalent token economies in which participants must accumulate multiple tokens before being able to exchange for a backup reinforcer. Lastly, the authors were also unable to account for the variability in the results of Jeffrey’s assessments, despite making efforts to avoid extraneous influence.

​

The implications of the study suggested that tokens can maintain behaviour even in situations where primary reinforcers are delayed. The authors suggested that directly delivering highly preferred reinforcers may be more effective than tokens when the objective is to maintain high response rates of a particular behaviour. Overall, Bonfonte et al. (2020) demonstrated the value of tokens as conditioned reinforcers and provided implications to research the effects of increasing the delay between token-production and token-exchange, investigate a variety of exchange ratios, and study generalized conditioned reinforcers in relation to primary reinforcers in token economies. 

 

​

Citation:

 

Bonfonte, S. A., Bourret, J. C., & Lloveras, L. A. (2020). Comparing the reinforcing efficacy of tokens and primary reinforcers. Journal of Applied Behavior Analysis. Advance online publication. https://doi.org/10.1002/jaba.675